No one believes that maintaining knowledge-based applications is easy, but perhaps it could be made less difficult. With the aid of screen shots, Nordlander, Freuder, and Wallace describe how an end user can modify, delete, and add constraint rules in a constraint-based reasoning system designed to support hospital inventory management. These maintenance actions become available when an order request violates policy or physical space constraints. For example, an order for too many items might violate a storage constraint in an operating theater. An end user, such as a head nurse, can then either reduce the quantity ordered or modify the constraint rule to reflect recently added storage capacity of which the system was unaware. End users may perform maintenance actions only if they have the same or higher access level as the original creator of the constraint rule. A key advantage of the system outlined is that maintenance and routine product ordering are interwoven, allowing maintenance on the fly.

From a human-computer interaction (HCI) perspective, however, I am not convinced that on-the-fly maintenance really works as portrayed. Nothing is mentioned about interface prototyping sessions, usability lab sessions, or satisfied usability requirements, as part of customer acceptance. The frequency and nature of maintenance actions by real end users is unknown, as the authors do not report on field sessions.

Written in a proof-of-concept style, this paper disappoints. With the system already handed over to the hospital, it misses the opportunity to report on all kinds of detailed evaluation work. Nevertheless, I recommend this paper to those researching the maintenance of knowledge-based applications.